Exploring Cell-Types in UNS data

The cell annotated as UNS by Cano-Gamez et al. were T cells that were neither stimulated by cytokines nor had an activated TCR. To facilitate the analysis, these cells were analyzed seperatly.

Loading the necessary libraries

import scanpy as sc
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt

# settings can be adapted individually
sc.settings.verbosity = 3            
sc.logging.print_header()             
sc.settings.set_figure_params(dpi = 100, format = 'png')

scanpy==1.7.2 anndata==0.7.6 umap==0.4.6 numpy==1.20.1 scipy==1.6.2 pandas==1.2.4 scikit-learn==0.24.2 statsmodels==0.12.2 python-igraph==0.9.1 louvain==0.7.0

Load preprocessed scRNA-seq data
See notebook "Data preprocessing" for this analysis part

canogamez = sc.read_h5ad("result_files/canogamez_preprocessing.h5ad") # change to your data path 

# create a path to store the preprocessed file
results_file = '/canogamez_UNS.h5ad' # change to your data path 

Dimension Reduction

PCA

sc.tl.pca(canogamez, svd_solver = 'arpack')

computing PCA
    on highly variable genes
    with n_comps=50
    finished (0:00:02)

Separate the data in UNS and stimulated

# select all UNS cells 
resting = canogamez.obs['cytokine.condition'] == 'UNS'

# create AnnData consisting only of UNS cells
canogamez_uns = canogamez[resting,:]

Plot PCA

fig, (ax1, ax2, ax3) = plt.subplots(1,3, figsize = (20,4), gridspec_kw = {'wspace':1})
ax1_dict = sc.pl.pca(canogamez_uns, color = 'cell.type', ax = ax1, show = False, annotate_var_explained = True)
ax2_dict = sc.pl.pca(canogamez_uns, color = 'cytokine.condition', ax = ax2, 
                     show = False,annotate_var_explained = True)
ax3_dict = sc.pl.pca(canogamez_uns, color = 'donor.id', ax = ax3, show = False, annotate_var_explained = True)

Trying to set attribute `.uns` of view, copying.

UMAP

sc.pp.neighbors(canogamez_uns, n_neighbors = 10, n_pcs = 40)
sc.tl.umap(canogamez_uns)

computing neighbors
    using 'X_pca' with n_pcs = 40
    finished: added to `.uns['neighbors']`
    `.obsp['distances']`, distances for each pair of neighbors
    `.obsp['connectivities']`, weighted adjacency matrix (0:00:01)
computing UMAP
    finished: added
    'X_umap', UMAP coordinates (adata.obsm) (0:00:10)

sc.pl.umap(canogamez_uns, color = ['cytokine.condition', 'cell.type'])

Louvain Clustering

The Louvain algorthim was chosen for clustering as it was used by Cano-Gamez et al.

! Caution: Rerunning your code will change the cluster composition due to randomness of the algorthim !

sc.tl.louvain(canogamez_uns, key_added = "louvain_1.0", random_state = 1)

running Louvain clustering
    using the "louvain" package of Traag (2017)
    finished: found 8 clusters and added
    'louvain_1.0', the cluster labels (adata.obs, categorical) (0:00:00)

sc.pl.umap(canogamez_uns, color = ['louvain_1.0', 'cell.type']) 

Explore the composition of the clusters

def count_pie(anndata, clustering, category):

    """generates a data frame with counts for a specific category within  
       the clusters and, plots values as pie chart"""

    # generate data frame with information for cluster
    clusters_df = anndata.obs[str(clustering)].to_frame()
    clusters_df[str(category)] = anndata.obs[str(category)]

    # generate empty dataframe for counted values
    number_clusters = len(np.unique(anndata.obs[str(clustering)]))
    row_names = list(np.unique(anndata.obs[str(clustering)]))
    row_names_long = ['cluster ' + name for name in row_names]
    col_names = list(anndata.obs[str(category)].cat.categories)
    df_cell_count = pd.DataFrame(0, columns=col_names,
    index=row_names_long)

    # fill dataframe with counts of the given categorie
    for i in range(0, number_clusters):
        cluster = clusters_df[str(clustering)] == str(i)
        cells_cluster = clusters_df[cluster]
        count_cells = cells_cluster.value_counts()
        for ic, vc in count_cells.items():
            df_cell_count.at['cluster ' + ic[0], ic[1]] = vc

    # plot as piechart
    from natsort import natsorted

    df_cell_count_T = df_cell_count.T
    df_cell_count_T
    df_cell_count_T.reindex(natsorted(df_cell_count_T.columns, ), axis=1)

    amount_plots = len(df_cell_count)
    amount_cols = 4
    amount_rows = int(np.ceil(amount_plots / amount_cols))
    fig, axes = plt.subplots(nrows=amount_rows, ncols=amount_cols,
                figsize=(15, 15))
    fig.tight_layout()

    for index, column in enumerate(df_cell_count_T):
        current_ax = axes[index // amount_cols, index % amount_cols]
        current_ax.set_title('{}'.format(column))
        current_data = df_cell_count_T[column]
        current_labels = list(current_data.index)
        current_data = list(current_data)
        current_ax.pie(current_data, labels=current_labels,
        autopct='%1.1f%%', startangle=90)
        current_ax.axis('equal')

    return df_cell_count, plt.show()

count_pie(canogamez_uns, 'louvain_1.0', 'cell.type')

(           Memory  Naive
 cluster 0     201   1171
 cluster 1    1243     66
 cluster 2     174    697
 cluster 3     638     47
 cluster 4     578     46
 cluster 5     122    117
 cluster 6      98     10
 cluster 7      56      5,
 None)

Most clusters can be clearly assigned to one particluar cell type. Only cluster 5 seems to conist of a mixture of memory and naive cells.

Rank genes

sc.tl.rank_genes_groups(canogamez_uns, groupby = 'louvain_1.0', method = 'wilcoxon', use_raw=True)

ranking genes
    finished: added to `.uns['rank_genes_groups']`
    'names', sorted np.recarray to be indexed by group ids
    'scores', sorted np.recarray to be indexed by group ids
    'logfoldchanges', sorted np.recarray to be indexed by group ids
    'pvals', sorted np.recarray to be indexed by group ids
    'pvals_adj', sorted np.recarray to be indexed by group ids (0:00:04)

sc.tl.dendrogram(canogamez_uns, groupby = 'louvain_1.0')

    using 'X_pca' with n_pcs = 50
Storing dendrogram info using `.uns['dendrogram_louvain_1.0']`

sc.pl.rank_genes_groups_matrixplot(canogamez_uns, n_genes = 3,cmap = 'bwr',
                                   standard_scale = "var", values_to_plot = 'scores')

Mostly ribosomal proteins. No indication of cell types

Cluster Annotation

Annotate naive cells

sc.pl.umap(canogamez_uns, color = ['louvain_1.0', 'cell.type'], legend_loc = 'on data') 

• Cluster 0 and 2 seem to represent the naive cells
  • similar gene expression
  • about 80 % naive cells
• Cluster 7 mostly memory according annotation cell.type but similar expression as 0 and 2 -> assign to Tn
• other clusters have to be assigned to nTreg, TCM, TEM and TEMRA as was done in the paper

Used marker genes

The marker genes from literature are based on the following website: https://www.biocompare.com/Editorial-Articles/569888-A-Guide-to-T-Cell-Markers/

T cell type / differentiation state	Marker genes mentioned in Cano-Gamez et al.
central memory T cells (Tcm)	PASK
effector memory T cells (Tem)	IL7R, KLRB1, TNFSF13B
terminally differentiated effector cells (TEMRA)	CCL4, GZMA
natural T regulatory cells (nTreg)	FOXP3, CTLA4

T cell type / differentiation state	Marker genes literature
T naive (Tn)	CCR7
central memory T cells (Tcm)	FAS, IL2RB, PRDM1
effector memory T cells (Tem)	CXCR3, ITGAL, CCR5, TBX21

Map marker genes on clusters

TEMRA

sc.pl.umap(canogamez_uns, color = ['louvain_1.0', 'GZMA', 'CCL4']) 

-> markers clearly expressed in cluster 5

nTreg

sc.pl.umap(canogamez_uns, color = ['louvain_1.0', 'FOXP3', 'CTLA4'])

-> marker genes expressed in cluster 6

Tn

sc.pl.umap(canogamez_uns, color = ['louvain_1.0',  'CCR7'])

-> marker genes mostly expressed in cluster 0, 2 and 7

Tcm

sc.pl.umap(canogamez_uns, color = ['louvain_1.0', 'FAS', 'IL2RB', 'PRDM1', 'PASK'])

-> Tcm marker expression present in cluster 3 and 1

TEM literature

sc.pl.umap(canogamez_uns, color = ['louvain_1.0', 'CXCR3', 'ITGAL', 'CCR5','TBX21'])

TEM paper

sc.pl.umap(canogamez_uns, color = ['louvain_1.0', 'IL7R', 'KLRB1', 'TNFSF13B'])

-> Tem cells can be assinged to cluster 4

Add annotation to AnnData

Create annotation

# adjust to individually identified clusters 
cluster_annotation = {
     '0': 'TN',
     '1': 'TCM',
     '2': 'TN',
     '3': 'TCM',
     '4': 'TEM',
     '5': 'TEMRA',
     '6': 'nTreg',
     '7': 'TN'
}

Add annotation to data

canogamez_uns.obs['cell type'] = canogamez_uns.obs['louvain_1.0'].map(cluster_annotation).astype('category')

Plotting

sc.pl.umap(canogamez_uns, color = 'cell type', legend_loc = 'on data', title = 'Annotated UNS cells',
           frameon = False, legend_fontsize = 10)

Similar annotation as by Cano-Gamez et al., see: https://cytokines.cellgeni.sanger.ac.uk/resting

Save data

canogamez_uns.write(results_file)